The project mainly aims to develop an autonomous car using a Unity 3D environment and complete the NHTSA Levels of Automation one by one. See Roadmap and Technologies for more detail.
-
Simple Unity 3D Environment
- The environment should be suitable for object detection tasks.
- Car controls must be realistic enough to simulate an autonomous vehicle.
- Integration with stereo camera sensors for realistic input data.
-
Road Segmantation and Lane Detection
-
Model Training with Reinforcement Learning for Level 1 Vehicle (Steering) Assist.
-
3D Object Detection Model (KITTI Dataset)
- Train a 3D object detection model using the KITTI dataset.
- The autonomous car should rely on stereo cameras (no LiDAR).
- Apply the trained model to detect and localize objects in the Unity environment.
-
Bird’s Eye View (BEV) Transformer System
- Implement a BEV Transformer System for 3D space prediction.
- Utilize stereo vision to predict the 3D structure of the environment.
- Explore combining the stereo vision system with more advanced models for enhanced accuracy.
-
Level 2 Vehicle Assist
-
Rule-based System Design for Realistic Traffic Ride
-
Level 3 Vehicle Assist
-
Level 4 Vehicle (Parking in Selected Parking Areas and Highway Riding) Assist
-
Level 5 Vehicle (Currently Dreaming) Assist
See the detailed documentation.
See my notes.
See the LICENSE
Screenshots will be updated.
Technologies will be updated.
Installation will be updated.
Local Run will be updated.
For support, email orhun868@gmail.com.
All contributions are welcome.
See CONTRIBUTING for ways to get started.
Please adhere to this project's CODE OF CONDUCT.
Appendix will be updated.